Add a Remove method to the Queue base class

Add a Remove method to the Queue base class

[Stefano Avallone, 2016-04-26 16:05:57]

This patch adds a Remove method to the Queue base class. This is needed by
fq-codel to drop the packet "from the head of the queue with the largest current
byte count" (as stated in draft-ietf-aqm-fq-codel-06).

Calling CoDel::Dequeue might drop some packets in addition to dequeuing a
packet, so it is not an option.

This patch depends on the fixes to bugs #2384 and #2389. All the changes can be
also seen via github:

https://github.com/stavallo/ns-3-dev-git/commits/tc-fixes

[Tom Henderson, 2016-05-05 21:58:45]

https://codereview.appspot.com/292540043/diff/1/src/network/utils/queue.h
File src/network/utils/queue.h (right):

https://codereview.appspot.com/292540043/diff/1/src/network/utils/queue.h#new...
src/network/utils/queue.h:75: bool Remove (void);
why not keep symmetry in the return value with Dequeue ()?

[Stefano Avallone, 2016-05-06 16:57:19]

https://codereview.appspot.com/292540043/diff/1/src/network/utils/queue.h
File src/network/utils/queue.h (right):

https://codereview.appspot.com/292540043/diff/1/src/network/utils/queue.h#new...
src/network/utils/queue.h:75: bool Remove (void);
basically because Remove is meant to drop a packet and the user should have no
interest in the dropped packet (otherwise, the user should call Dequeue).
Anyway, if you prefer to keep symmetry, we can have Remove return the dropped
packet.

[Tom Henderson, 2016-05-09 17:15:51]

On 2016/05/06 16:57:19, Stefano Avallone wrote:
> https://codereview.appspot.com/292540043/diff/1/src/network/utils/queue.h
> File src/network/utils/queue.h (right):
> 
>
https://codereview.appspot.com/292540043/diff/1/src/network/utils/queue.h#new...
> src/network/utils/queue.h:75: bool Remove (void);
> basically because Remove is meant to drop a packet and the user should have no
> interest in the dropped packet (otherwise, the user should call Dequeue).
> Anyway, if you prefer to keep symmetry, we can have Remove return the dropped
> packet.

I don't feel strongly but it just seems to me that you can 1) keep symmetry, and
2) provide more information to the caller with no cost to us (a null pointer
value is equivalent to false).  Who knows whether someone in the future may find
it useful to inspect the removed packet?

[Stefano Avallone, 2016-05-11 10:38:39]

Modified as suggested by Tom.

[Stefano Avallone, 2016-05-17 13:46:37]

Ok to commit? Thanks.

[Tom Henderson, 2016-05-17 22:38:58]

On 2016/05/17 13:46:37, Stefano Avallone wrote:
> Ok to commit? Thanks.

OK by me.